Professional Documents
Culture Documents
ALLURI Et Al. Large-Scale Brain Networks Emerge From Dynamic Processing of Musical Timbre, Key and Rhythm (Tango Brain) PDF
ALLURI Et Al. Large-Scale Brain Networks Emerge From Dynamic Processing of Musical Timbre, Key and Rhythm (Tango Brain) PDF
NeuroImage
journal homepage: www.elsevier.com/locate/ynimg
Large-scale brain networks emerge from dynamic processing of musical timbre, key
and rhythm
Vinoo Alluri a,⁎, Petri Toiviainen a, 1, Iiro P. Jääskeläinen c, d, 2, Enrico Glerean c, d, 2,
Mikko Sams c, d, 2, Elvira Brattico a, b, 3
a
Finnish Centre of Excellence in Interdisciplinary Music Research, University of Jyväskylä, Finland
b
Cognitive Brain Research Unit, Institute of Behavioral Sciences, University of Helsinki, Finland
c
Mind and Brain Laboratory, Department of Biomedical Engineering and Computational Science (BECS), Aalto University School of Science, Finland
d
Advanced Magnetic Imaging Centre, Aalto University School of Science, Finland
a r t i c l e i n f o a b s t r a c t
Article history: We investigated the neural underpinnings of timbral, tonal, and rhythmic features of a naturalistic musical stimu-
Received 21 June 2011 lus. Participants were scanned with functional Magnetic Resonance Imaging (fMRI) while listening to a stimulus
Revised 13 October 2011 with a rich musical structure, a modern tango. We correlated temporal evolutions of timbral, tonal, and rhythmic
Accepted 6 November 2011
features of the stimulus, extracted using acoustic feature extraction procedures, with the fMRI time series. Results
Available online 12 November 2011
corroborate those obtained with controlled stimuli in previous studies and highlight additional areas recruited
Keywords:
during musical feature processing. While timbral feature processing was associated with activations in cognitive
Music processing areas of the cerebellum, and sensory and default mode network cerebrocortical areas, musical pulse and tonality
fMRI processing recruited cortical and subcortical cognitive, motor and emotion-related circuits. In sum, by combining
Computational feature extraction neuroimaging, acoustic feature extraction and behavioral methods, we revealed the large-scale cognitive, motor
Naturalistic stimulus and limbic brain circuitry dedicated to acoustic feature processing during listening to a naturalistic stimulus. In ad-
Temporal evolution dition to these novel findings, our study has practical relevance as it provides a powerful means to localize neural
processing of individual acoustical features, be it those of music, speech, or soundscapes, in ecological settings.
© 2011 Elsevier Inc. All rights reserved.
Introduction instrument's timbre from that of another, perceive the leading melody
by extracting pitch height, and feel the beat (Bregman, 1990; Janata et
Music is fundamental to humans across all cultures and is capable of al., 2002a, 2002b). In this process, domain-specific neural mechanisms
eliciting intense emotions (Salimpoor et al., 2011). Uncovering the neu- for acoustic feature analysis and integration as well as domain-general
ral underpinnings of music processing has become a central theme in neural circuits of attention and memory are required. In particular, hier-
cognitive neuroscience in the past decade, as evidenced by the con- archical processing within the auditory cortex going from more simple
stantly increasing corpus of studies on this topic. The intrinsically multi- to more complex features (Chevillet et al., 2011; Patterson et al., 2002),
dimensional nature of music renders this task challenging. More and hemispheric specialization (Samson et al., 2011; Zatorre et al.,
specifically, music comprises several perceivable features of varying 2002) for spectral vs. temporal acoustic variations, have been identified
levels of abstraction, such as loudness, pitch (the organization of sounds as putative principles of functional organization of acoustic feature-
along a scale from low to high), rhythm (the perceptual organization of related processing. Previous neuroimaging studies of music have
sound events in time) and timbre (property that allows to distinguish attempted to identify brain structures involved in the perception of
between different instrument sounds having the same pitch and loud- music-related perceptual features, such as pitch (Patterson et al.,
ness). Perceiving polyphonic music involves automatic segregation of 2002), sensory dissonance (Blood et al., 1999; Koelsch et al., 2006),
the musical information in the brain. For instance, when listening to a rhythm (Chen et al., 2008; Grahn and Rowe, 2009), timbre (Caclin et
piece of music played by an orchestra we are able to distinguish one al., 2006; Halpern et al., 2004), and key (Janata et al., 2002a, 2002b).
However, while these studies have successfully identified brain regions
participating in processing of individual musical features they have re-
⁎ Corresponding author at: Department of Music, University of Jyväskylä, PL 35(M), lied on controlled auditory paradigms in which these features have
40014 Jyväskylä, Finland. been presented in isolation and manipulated artificially. Although a
E-mail address: vinoo.alluri@jyu.fi (V. Alluri). few studies have investigated brain responses during continuous listen-
1
Department of Music, University of Jyväskylä, PL 35(M), 40014 Jyväskylä, Finland.
2
BECS, Dept. of Biomedical Engineering and Computational Science, P.O.Box 12200,
ing to relatively simple musical stimuli (Janata et al., 2002a, 2002b;
FI-00076 Aalto, Finland. Schaefer et al., 2009), it has not been previously studied how the
3
Institute of Behavioral Sciences, P.O.B. 9, 00014, University of Helsinki, Finland. human brain processes, in parallel, the multitude of musical features
1053-8119/$ – see front matter © 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2011.11.019
3678 V. Alluri et al. / NeuroImage 59 (2012) 3677–3689
when participants are listening to a record of real orchestra music dur- processing, such as the rostromedial prefrontal cortex (Janata et al.,
ing neuroimaging. 2002a, 2002b). In addition, as tonality processing draws on long-
In the visual modality, recent evidence suggests that the brain pro- term knowledge of hierarchical tonality structures (Krumhansl,
cesses visual stimuli presented in a more ecological setting differently 1990), we expected the brain areas related to memory processing,
than when presented in conventional controlled settings (Hasson et such as the hippocampus (see Burianova et al., 2010 for an overview)
al., 2004). Assuming that this finding is generalizable across sensory to be activated. We hypothesized rhythm-related features to recruit,
modalities, one could expect that the majority of studies in the audi- in addition to areas in the auditory cortex, cortical and subcortical
tory modality, in which acoustic features were artificially manipulat- areas related to motor processing, such as the premotor and supple-
ed, may have revealed an incomplete picture of brain function related mentary motor areas, and subcortical structures involved in the pro-
to musical feature processing. Therefore, studying music listening as a cessing of time intervals such as the basal ganglia (Harrington et al.,
continuous process using naturalistic stimuli could provide more ac- 1998; Janata and Grafton, 2003; Rao et al., 2001; Schwartze et al.,
curate accounts of the processing of musical features in the brain. 2011). Furthermore, as tonal and rhythmic features are known to elicit
We employed a stimulus-wise and task-wise more ecological setting expectations in listeners (Janata, 2005; Zanto et al., 2006), we hypoth-
in which participants freely listened to real music without performing esized them to shape activations in the higher-order areas in the brain,
any other task, in order to determine the neural mechanisms and struc- such as the supplementary motor areas, which are known to be in-
tures responsible for musical feature processing under realistic condi- volved in perceptual tasks having an anticipatory component
tions. To tackle the complexity of the problem, we introduced a novel (Schubotz and von Cramon, 2002).
interdisciplinary approach combining neuroimaging with computa-
tional acoustic feature extraction and behavioral psychology. As music Materials and methods
stimulus we chose the modern tango Adios Nonino by Astor Piazzolla.
The participants were scanned with fMRI while listening to this piece. Participants
Temporal evolutions of acoustic components representing timbral,
tonal and rhythmic features of the stimulus were computationally Eleven healthy participants (with no neurological, hearing or psy-
extracted and validated via a perceptual experiment. Following this chological problems) with formal musical training participated in the
we performed correlation analyses of the time series of the individual study (mean age: 23.2 ± 3.7 SD; 5 females). We chose participants
acoustic components and the time series of the BOLD signal. with formal musical training as it has been shown that musicians dis-
In light of previous studies (Samson et al., 2011), we hypothesized play stronger neural responses to various musical features in compari-
that timbral components would activate mainly sensory areas, such as son to non-musicians (Pantev et al., 2001; Wong et al., 2007). Five
the superior temporal gyrus (STG) and the Heschl's gyrus (HG). More- participants were educated in and performed mainly classical music,
over, we expected that the spectrally varying timbral components two musicians were trained in folk and jazz music, and the rest were
would activate particularly the caudolateral and anterior superior playing mainly pop/rock music. Four musicians played string instru-
temporal regions, respectively (see Samson et al., 2011 for an over- ments, three percussive instruments, two wind instruments, and two
view). In addition, we predicted hemispheric lateralization, in partic- keyboard instruments. All participants, except one, were also able to
ular, that right hemispheric regions would show larger areas play other instruments along with their main one. These musicians
involved in the processing of these features (Zatorre et al., 2002). Pro- started to play their main instrument on average at 9.1 ± 3.4 SD years
cessing of tonality-related features was expected to recruit areas in of age, and their second instrument at 10.5 ± 3.7 SD years, collecting a
the brain formerly known to be neural substrates of tonality total amount of years of training equal, on average, to 16.1 ± 6 SD.
Fig. 1. Loadings of the features on the first 9 Principal Components (PC) as a result of principal component analysis with varimax rotation. The x-axis indicates principal component
number in order of decreasing amount of variance explained. The shade of gray indicates the respective loading with white indicating the maximum and black the minimum. PCs 4, 5,
and 8 are not labeled as they failed to correlate with the perceptual ratings and thus were excluded from subsequent analyses.
V. Alluri et al. / NeuroImage 59 (2012) 3677–3689 3679
Fullness .87 19
The stimulus used in the experiment was the tango Adios Nonino by .28
Astor Piazzolla of a duration of 8 min and 32 s. This piece of music was Brightness .27 .87 20
Timbral complexity .34 .90 18
selected due to its high range of variation in several musical features
Rhythmic complexity .30 .90 21
such as dynamics, timbre, tonality and rhythm, while having an appro- Key clarity .54 .96 21
priate duration for the experimental setting used. Pulse clarity .55 .96 21
Event synchronicity .25 .85 17
fMRI measurements Activity .61 .97 20
Dissonance .55 .96 19
Music Information Retrieval (MIR) (Tzanetakis and Cook, 2002). These ⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.
3680 V. Alluri et al. / NeuroImage 59 (2012) 3677–3689
Fig. 2. Selected slices depicting the areas of the brain that displayed significant mean inter-subject correlation (p b .001).
As visible in Fig. 1, the first PC has the highest loadings from Sub- Procedure
bands No. 2 and No. 3 Flux. The presence of spectral fluctuations in The participants were given written instructions, following which
lower bands of the spectrum has been associated previously to perceptu- the listening experiment took place in a silent room. To present the
al ‘Fullness’ (Alluri and Toiviainen, 2010). The second PC, with the high- stimuli and obtain the ratings, a graphical interface was developed
est loadings from spectral centroid, spectral roll-off, high energy–low in Max/MSP. The experiment was divided into nine sections, in each
energy ratio and zero crosssing rate measures the perceptual ‘Brightness’ of which the participants were asked to rate the stimuli according
(Alluri and Toiviainen, 2010, 2012). PC3 represents the spread and to one of the scales. Each scale was divided into 9 levels with the ex-
flatness (Wiener entropy) of the spectrum. In perceptual terms, it can tremes of the scale indicating low and high values of the perceptual
be regarded to represent timbral complexity. As can further be seen scale (e.g: 1 = Low Pulse Clarity, 9 = High Pulse Clarity), from
from Fig. 1, each of the PCs from 4 through 8 has high contributions which the subject could choose the level that best described the
from only one of the long-term features representing ‘Mode’, ‘Rhythmic music excerpt presented. The participants were able to listen to
Complexity’, ‘Key Clarity’, ‘Pulse Clarity’, and ‘Event Synchronicity’ re- each excerpt as many times as they wished. The order of presentation
spectively (see Appendix A for details). PC 9 has the highest loadings of each of the sections and the music excerpts in each section was
from roughness and flux of the high end of the spectrum. This can be randomized. Prior to the actual experiment, the participants were
regarded to represent ‘Activity’ (Alluri and Toiviainen, 2010, 2012). allowed to familiarize themselves with the working of the interface.
The loudness feature of the stimulus, characterized by RMS energy, failed
to load highly onto any of these PCs suggesting that none of the PCs cor- Results
related notably with RMS energy. The perceptual ratings were initially checked for inconsistencies and
To validate the perceptual relevance of the labels given to the outliers. First, for each scale, the participants with a negative mean
acoustic components, we conducted a perceptual listening experi- inter-subject correlation were eliminated. Following this, participants
ment where participants were asked to rate selected segments repre- with a mean inter-subject correlation two SDs below the overall mean
senting varying levels of the acoustic components on various inter-subject correlation were eliminated. As a result, one to four partic-
perceptual scales. The procedure is explained in detail as follows. ipants were excluded for each scale. As visible in Table 1, Cronbach's
alpha revealed high agreement between the participants' ratings, indi-
Perceptual experiment cating fairly consistent opinions among listeners with respect to the
perceptual scales. Therefore, for subsequent analysis, the individual rat-
Stimuli ings for each scale were averaged across the participants. As can be seen
We selected nine sets of stimuli comprising 30 6-second segments from Table 2, significant correlations were observed between the rating
each, obtained from the stimulus used in the fMRI experiment. For scales of Fullness, Brightness, Timbral Complexity, Key Clarity, Pulse
each of the sets, the excerpts were chosen to represent varying levels Clarity and Activity and the respective acoustic components. The
of one of the acoustic components. The acoustic component repre- remaining scales of Rhythmic Complexity and Event Synchronicity
senting mode was excluded, because according to initial analysis of failed to correlate significantly with the corresponding acoustic compo-
the fMRI data it failed to display any significantly correlating areas. nents and were hence excluded from further analyses. Therefore, the
Additionally, we wanted to limit the duration of the experiment to temporal evolutions of only the corresponding principal component
avoid fatigue effects. To obtain the excerpts, for each acoustic compo- scores of these six PCs were used for further analysis and will hence-
nent, the raw acoustic features were multiplied by the respective PC forth be referred to as acoustic components.
loadings. Subsequently, we averaged the hence obtained PC score
time series within 6-second moving windows using a hop-size of Statistical analysis
1 s. Next, each of these averaged time series were rank ordered and
sampled (n = 30) equidistantly in order to obtain segments that Initially, the inter-subject consistency of the fMRI data was checked
represented varying levels of the respective PC scores and spanned using mean inter-subject correlation as a measure. The significance of
the entire range of variation. the correlations was estimated by means of a Monte Carlo approach.
Following this, we employed the correlation analysis approach utilized
Participants by He et al. (2008). First, for each participant, the Pearson correlation
Twenty-one musicians (15 females, age M = 23.7 years, SD= 3.5) coefficients (r) were obtained per voxel and per acoustic component.
participated in rating the music excerpts. All reported having formal The r maps were then converted to respective Z-score maps using Fish-
music education (M= 15.8 years, SD= 6.1 years, minimum= 3 years). er's Z transformation, normalized by the factor 1 / √(df− 3), where df
They practiced on average 11.2 (SD = 9) hours per week. Seven represents the estimated number of effective degrees of freedom. The
reported familiarity with the piece Adios Nonino. All of the participants effective degrees of freedom were estimated using the approach de-
reported listening to music the average being 12 h/week. None of scribed by Pyper and Peterman (1998) (see Appendix B for details).
these participants took part in the fMRI experiment. However, the two As a result, the effective degrees of freedom varied from a minimum
groups shared similar demographic properties in terms of their ages of 29 (correction factor = 7.97) to a maximum of 68 (correction fac-
as well as their musical and cultural backgrounds. tor= 3.40) across the acoustic components.
V. Alluri et al. / NeuroImage 59 (2012) 3677–3689 3681
For each acoustic component, in order to obtain the group maps The group maps hence obtained for each component were thre-
we employed the combining tests procedure described by Lazar sholded at a significance level of p b .001. Following this, in order to min-
(2008). In this approach, individual Z-score maps were first con- imize Type I errors, we corrected for multiple comparisons using cluster
verted to p-maps and were then pooled using the Fisher's p-value size thresholding. To determine the thresholds for multiple compari-
technique (Fisher, 1950) to create the group maps. The individual sons correction, we performed a Monte Carlo simulation of the ap-
p-values of each voxel were pooled using the equation below to ob- proach described by Ledberg et al. (1998) (see Appendix B). As result,
tain a T-statistic (Eq. (1)) that is modeled as a Chi-square distribu- we obtained a cluster size threshold of 22 voxels for p b .001 (Z= 3.29).
tion with 2k degrees of freedom where k represents the number of
participants. Results
X
k Fig. 2 displays the results of the correlation analysis performed to
T ¼ −2 log pi : ð1Þ test the consistency between the participants' fMRI responses. As can
i¼1 be seen, relatively large areas of the brain were found to display
Fig. 3. Lateral and mid-sagittal views of the left and right hemispheres of the brain showing regions correlating significantly with timbral components. The significance threshold for
the correlations was set at p = .001. Cluster correction was performed at a significance level of p = .001 (Z = 3.29) corresponding to a cluster size of 22 voxels. The exact Z-values are
listed in Table 3. The areas indicated in red and blue correspond to the brain areas that correlated positively and negatively to each of the timbral components, respectively.
3682 V. Alluri et al. / NeuroImage 59 (2012) 3677–3689
Table 3
Correlation results between acoustic components and brain activity using a significance threshold p b .001. The clusters were obtained using the 18-connectivity scheme employed
in SPM. The significant voxels were cluster corrected at Z > 3.29 (p b .001 and cluster size > 22). The coordinates are in Talairach space and indicate the location of the global max-
imum within each cluster. For each cluster, ROI analysis was performed using MarsBar. Each cluster is numbered and named after the ROI containing the voxel with the maximal Z-
value. The sub-clusters obtained from the ROI analysis are displayed under each main cluster. The coordinates for the sub-clusters are in Talairach space and indicate the location of
the local maximum within each cluster.
(A) Fullness
Positive correlation
1. Superior temporal gyrus 22 2325 51 − 14 1 7.35 5. Superior temporal gyrus 22 2283 − 50 − 21 3 7.34
Superior temporal gyrus 1646 51 − 14 1 7.35 Superior temporal gyrus 1137 − 50 − 21 3 7.34
Heschl's gyrus 203 50 − 15 4 6.82 Heschl's gyrus 80 − 53 − 15 8 5.20
Insula 17 50 −4 0 4.70 Middle temporal gyrus 890 − 50 − 19 1 6.94
Postcentral gyrus 8 63 − 15 14 3.78
Rolandic operculum 193 40 − 28 16 6.13 Rolandic operculum 64 − 44 − 28 14 6.23
Supramarginal gyrus 100 50 − 34 24 4.96 Supramarginal gyrus 13 − 51 − 25 14 4.71
Superior temporal pole 20 53 2 −5 4.38
Cerebellum
2. Inferior semi-lunar lobule 76 30 − 78 − 35 4.25
VIIB 1 26 − 78 − 36 3.58
Crus II 75 30 − 78 − 35 4.25
3. Declive 39 32 − 63 − 17 3.76 6. Declive 39 − 14 − 73 − 13 3.75
VI 39 32 − 63 − 17 3.76 VI 39 − 14 − 73 − 13 3.75
4. Tonsil 40 6 − 45 − 38 3.60
IX 23 16 − 47 − 40 3.41
7. Uvula 79 − 28 − 63 − 24 4.33
VI 28 − 28 − 61 − 24 4.18
Crus I 50 − 28 − 63 − 24 4.33
VIIB 22 −8 − 71 − 30 3.95
VIII 2 −8 − 69 − 30 3.50
8. Pyramis 66 − 28 − 68 − 34 4.10
VIIB 10 − 28 − 68 − 34 4.10
VIII 2 − 26 − 66 − 34 3.45
Crus II 54 − 30 − 70 − 34 4.09
9. Pyramis 23 − 20 − 83 − 31 4.01
Crus II 23 − 20 − 83 − 31 4.01
Negative correlation
1. Postcentral gyrus 2 67 44 − 23 49 3.66 3. Postcentral gyrus 3 191 − 36 − 31 48 4.19
Postcentral gyrus 67 44 − 23 49 3.66 Postcentral gyrus 83 − 36 − 31 48 4.19
2. Inferior parietal lobule 40 31 40 − 38 55 3.81 Inferior parietal gyrus 95 − 32 − 41 43 3.70
Inferior parietal gyrus 12 40 − 38 52 3.41
Superior parietal gyrus 9 40 − 38 53 3.66
Postcentral gyrus 10 40 − 38 55 3.81
4. Middle frontal gyrus 11 27 − 26 38 − 19 4.09
Inferior orbitofrontal gyrus 11 − 30 42 − 17 3.54
5. Superior parietal lobule 7 165 − 18 − 61 56 3.71
Superior parietal gyrus 75 − 18 − 61 56 3.71
Postcentral gyrus 18 − 24 − 42 56 3.58
Precuneus 72 − 16 − 61 58 3.53
(B) Brightness
Positive correlation
1. Superior temporal gyrus 22 3166 50 − 19 3 8.13 6. Superior temporal gyrus 22 3245 − 55 − 15 3 8.13
Superior temporal gyrus 2070 50 − 19 3 8.13 Superior temporal gyrus 1516 − 55 − 15 3 8.13
Heschl's gyrus 225 48 − 15 4 8.13 Heschl's gyrus 153 − 53 − 15 8 7.73
Insula 51 50 −4 0 6.16 Insula 13 − 46 −8 −1 4.50
Postcentral gyrus 27 63 − 15 14 5.01 Postcentral gyrus 3 − 63 − 21 16 3.41
Rolandic operculum 274 40 − 28 16 6.80 Rolandic operculum 109 − 46 − 28 14 6.89
Supramarginal gyrus 99 48 − 34 24 4.95 Supramarginal gyrus 16 − 63 − 23 14 5.37
Middle temporal gyrus 90 46 − 33 3 4.17 Middle temporal gyrus 1192 − 59 − 12 −3 8.13
Temporal pole 74 51 2 −7 5.90 Temporal pole 11 − 57 5 −9 4.40
2. Precentral Gyrus 6 64 55 0 41 4.05 7. Precentral gyrus 6 28 − 55 0 42 4.39
Precentral gyrus 53 55 0 41 4.05 Precentral gyrus 17 − 55 0 42 4.39
Middle frontal gyrus 11 51 −1 50 3.69 Postcentral gyrus 10 − 55 0 41 4.21
3. Putamen 38 24 11 −6 3.63
Putamen 36 24 11 −6 3.63
4. Putamen 30 14 10 1 3.38
Putamen/pallidum 7 14 6 −4 3.33
Putamen 10 22 12 5 3.26
Cerebellum
5. Declive 186 − 32 − 63 − 22 4.34
VI 102 − 28 − 59 − 22 4.28
CrusI 84 − 32 − 63 − 22 4.34
Negative correlation
1. Medial frontal gyrus 9 26 − 20 42 15 3.68
Middle frontal gyrus 7 − 20 43 14 3.63
Superior frontal gyrus 2 − 20 42 18 3.31
2. Posterior cingulate 29 26 −6 − 44 10 3.50
Calcrine 4 −6 − 44 8 3.44
V. Alluri et al. / NeuroImage 59 (2012) 3677–3689 3683
Table 3 (continued)
Right hemisphere BA k x Y z Z-value Left hemisphere BA k x y z Z-value
(C) Activity
Positive correlation
1. Superior temporal gyrus 22 2368 51 − 14 1 7.92 5. Superior temporal gyrus 22 2378 − 50 − 21 3 7.48
Superior temporal gyrus 1632 51 − 14 1 7.92 Superior temporal gyrus 1185 − 50 − 21 3 7.48
Heschl's gyrus 211 50 − 15 4 7.31 Heschl's gyrus 91 − 53 − 15 8 5.58
Rolandic operculum 209 40 − 28 16 6.26 Rolandic operculum 73 − 44 − 28 14 5.87
Supramarginal gyrus 101 50 − 34 24 4.63 Supramarginal gyrus 11 − 53 − 25 14 4.36
Temporal pole 24 51 2 −5 4.67 Temporal pole 15 − 57 7 −9 3.78
Insula 24 50 −4 0 5.45 Middle temporal gyrus 893 − 50 − 19 1 7.09
Postcentral gyrus 8 63 − 15 14 3.87
2. Medial frontal gyrus 11 22 8 61 − 15 3.69
Medial frontal gyrus 22 8 61 − 15 3.69
Cerebellum
3. Inferior semi-lunar lobule 70 30 − 79 − 35 4.42 6. Uvula 91 − 30 − 63 − 24 4.13
VIIB 1 26 − 78 − 36 3.60 VI 19 − 28 − 61 − 24 3.96
CrusII 69 30 − 79 − 35 4.42 CrusI 72 − 30 − 63 − 24 4.13
4. Declive 50 32 − 63 − 15 3.89 7. Pyramis 106 − 32 − 70 − 32 4.11
VI 50 32 − 63 − 15 3.89 VIIB 8 − 28 − 68 − 34 3.86
VIII 1 − 26 − 66 − 34 3.40
CrusII 97 − 32 − 70 − 32 4.11
8. Declive 57 − 16 − 73 − 13 3.91
VI 57 − 16 − 73 − 13 3.91
Negative correlation
1. Postcentral gyrus 3 46 42 − 25 49 3.51 2. Postcentral gyrus 3 137 − 36 − 31 48 4.26
Postcentral gyrus 46 42 − 25 49 3.51 Postcentral gyrus 72 − 36 − 31 48 4.26
Inferior parietal lobule 56 − 38 − 35 44 3.64
3. Superior parietal lobule 7 105 − 18 − 53 58 3.61
Superior parietal lobule 46 − 18 − 53 58 3.61
Postcentral gyrus 9 − 24 − 42 56 3.43
Precuneus 50 − 16 − 53 58 3.47
4. Middle frontal gyrus 6 62 − 32 8 46 3.63
Middle frontal gyrus 24 − 32 8 46 3.63
Precentral gyrus 38 − 32 6 44 3.60
5. Superior frontal gyrus 9 24 − 20 52 25 3.48
Superior frontal gyrus 24 − 20 52 25 3.48
Middle frontal gyrus 19 − 22 52 27 3.48
Table 3 (continued)
Right hemisphere BA k x Y z Z-value Left hemisphere BA k x y z Z-value
Putamen 1 − 32 − 11 4 3.15
5. Cingulate gyrus 24 50 14 −6 44 3.66
Middle cingulate gyrus 27 14 −6 44 3.66
Supplementary motor area 22 14 −4 44 3.54
significant mean inter-subject correlations with the maximum found associated with increased activation in the declive, uvula, and pyra-
in the auditory cortices (r = .64, p b .0001). Following this, we per- mis (lobule VI, Crus I and II). Increase in the timbral components of
formed correlation analysis between the fMRI data and acoustic com- Brightness and Timbral Complexity was associated with increased ac-
ponents. As described in Perceptual experiment, perceptual tivation in the declive only.
validation of the acoustic components resulted in a compact set of Outside of the auditory cortex and the cerebellum, we observed cer-
six acoustic components that represented the temporal evolution of ebrocortical areas correlating negatively with the timbral components.
the main timbral, tonal, and rhythmic features in the stimulus. Tim- In particular, decreased Activity and Fullness was found to be associated
bral components comprised perceived Fullness, Activity, Brightness, with increased activations in the bilateral postcentral gyrus (BA 2 and
and Timbral complexity of the stimulus. Tonal and Rhythmic compo- 3), and the left precuneus (BA 7). In addition, low levels of Fullness
nents comprised Key Clarity and Pulse Clarity respectively (audio ex- were associated with increased activations in the bilateral inferior pari-
cerpts depicting examples of low levels and high levels of all six etal gyrus (BA 40), and those of Activity were associated with increased
acoustic components can be found as Supplementary material: activations in the left superior frontal gyrus (BA 9) and left medial fron-
AudiomaterialS1). First-level analysis comprised correlating these tal gyrus (BA 6). Increased activations in the right medial frontal gyrus
six components with fMRI time series at an individual level. Follow- (BA 11) were found to be associated with increasing Activity.
ing this, second-level analysis involved pooling individual results to Furthermore, increase in Brightness recruited the bilateral precen-
obtain group maps for each acoustic component. The results are dis- tral gyrus (BA 6), and the right putamen. Reduced levels of Brightness
cussed below. in the stimulus, on the other hand were associated with increased ac-
tivations in two left hemispheric clusters of the medial frontal gyrus
Timbral feature processing in the brain (BA 9) and the posterior cingulate cortex (BA 29). No negative corre-
lations were found for Timbral Complexity.
Correlation analyses revealed that the presence of high values
in all the timbral features namely Fullness, Brightness, Timbral
Complexity and Activity, was associated with increased neuronal Rhythmic and tonal feature processing in the brain
activation in the bilateral STG (BA 22) (see Fig. 3). Additionally, a
lateralization effect was found wherein the right hemisphere dis- High levels of Pulse Clarity were linked with increased activation
played positive correlations with timbral components in larger in the bilateral STG (BA 22), as well as the right primary auditory cor-
proportions of the HG, rolandic operculum, supramarginal gyrus, tex (BA 41) (see Fig. 4 and Table 3). In contrast, decreased Pulse Clar-
and superior temporal pole than the left hemisphere (see ity recruited the right inferior temporal gyrus (ITG, BA 37) and
Table 3). In contrast, the left hemispheric middle temporal gyrus precuneus (BA 7). In addition, decreased Pulse Clarity was associated
had a larger proportion displaying such a correlation than its with high levels of activation in several subcortical limbic areas in-
right hemispheric counterpart (see Table 3 for details). cluding the left hemispheric amygdala, hippocampus and putamen,
Another group of activations was found in the cerebellum. In par- the bilateral mid-cingulate gyrus (BA 32) in the vicinity of the supple-
ticular, high values of Fullness and Activity in the stimulus were mentary motor area, and the bilateral insula (BA 13).
V. Alluri et al. / NeuroImage 59 (2012) 3677–3689 3685
Fig. 4. Lateral and mid-sagittal views of the left and right hemispheres of the brain showing regions correlating significantly with Pulse Clarity and Key Clarity components. The
significance threshold for the correlations was set at p = .001. Cluster correction was performed at a significance level of p = .001 (Z = 3.29) corresponding to a cluster size of 22
voxels. The exact Z-values are listed in Table 3. The areas indicated in red and blue correspond to the brain areas that correlated positively and negatively to each of the components,
respectively.
The presence of unclear key (i.e., low levels of Key Clarity) was activations in subcortical emotion-related areas along with activations
found to be associated with increased activation in cortical and sub- in cognitive and somatomotor cerebrocortical areas.
cortical areas such as parts of the bilateral precentral gyrus (BA 3), Timbre-related acoustic components correlated positively with acti-
the right mid-cingulate gyrus (BA 24) in the vicinity of the supple- vations in large areas of the temporal lobe (STG, HG, and MTG) (see
mentary motor area and right postcentral (BA 6) gyrus, and the left Fig. 3). These areas were mostly activated during moments of high Full-
hemispheric superior frontal gyrus (BA 9), the left insula and the bi- ness, Activity, Brightness and Timbral Complexity, which were often as-
lateral rolandic operculum (see Fig. 4). No positive correlations sociated with quicker passages in the stimulus with several pitches
were observed for Key Clarity. usually played by several instruments. While there exists a dearth of
studies regarding neural correlates of timbre processing in a musical
Discussion context, or polyphonic timbre processing, evidence from neural studies
on monophonic timbre has repeatedly pointed at the involvement of
In the present study, we investigated the neural correlates of tim- the bilateral STG and HG (Caclin et al., 2006; Halpern et al., 2004). The
bral, tonal, and rhythmic feature processing of a naturalistic music stim- current findings suggest that the same brain areas are recruited in poly-
ulus. To this end we employed a novel paradigm combining phonic timbre processing. As hypothesized based on Samson et al.
neuroimaging, computational acoustic feature extraction and behavior- (2011), we found the caudolateral and anteriolateral parts of the STG,
al psychology. Participants were scanned using fMRI while they freely specifically in the right hemisphere to be involved in timbral feature
listened to the musical piece Adios Nonino by Astor Piazzolla. First, processing. Furthermore, this finding also supports our hypothesis of
inter-subject consistency on a voxel-by-voxel basis was evaluated interhemispheric specialization in the auditory cortices with regard to
using mean inter-subject correlation as a measure. Following this, the the processing of timbral features, with the right temporal lobe display-
evolution of musical features in the piece was obtained using sophisti- ing larger areas with significant correlations with these features.
cated acoustic feature extraction procedures. Based on a perceptual Negative correlations between timbral features of Activity and
test we selected a set of six acoustic components representing the Fullness and brain activity were observed in cerebrocortical regions
main timbral, tonal, and rhythmic features present in this piece. Follow- in the vicinity of the left superior frontal gyrus (BA 9), the left precu-
ing this, the neural underpinnings of these acoustic components were neus and surrounding parietal areas, and the ventral medial prefron-
investigated by correlating their time series with the time series of the tal cortex. These areas are known to be part of the default mode
BOLD signal. Overall, our results highlighted the brain structures re- network (DMN). The DMN is a neural circuit constantly monitoring
sponsible for the processing of an extensive set of timbral, tonal and the sensory environment and displaying high activity during lack of
rhythmic features. The results corroborate findings reported in previous focused attention on external events (Fox et al., 2009; McAvoy et
neuroimaging studies, which have used artificial and acoustically re- al., 2008). As low values in Activity and Fullness were mostly associ-
duced conditions to investigate musical feature processing. Moreover, ated with sections in the stimulus with sparse texture played by the
they also highlight additional brain structures involved in musical fea- piano, thereby resulting in lower levels of auditory–cognitive load,
ture processing. Timbral features activated mainly perceptual and the activation of the DMN during these moments is in line with pre-
resting-state or default mode areas of the cerebrum and cognitive vious results (Levitin and Menon, 2003; Pallesen et al., 2009; Uddin
areas of the cerebellum. In contrast, for tonal and rhythmic features, et al., 2009). In the visual modality, a network comprising several
we observed for the first time during listening to a naturalistic stimulus, parietal areas such as the precuneus and the supramarginal gyrus
3686 V. Alluri et al. / NeuroImage 59 (2012) 3677–3689
has been found to be highly activated with reduced visual attention or processing. However, further investigations are required to ascer-
load (Hahn et al., 2006). Similarly, the left posterior cingulate, one of tain this link. Koelsch et al. (2006) in their fMRI study contend
the central structures of the DMN, was observed to be deactivated that the rolandic operculum in conjunction with the anterior supe-
during moments in the stimulus with high Brightness, which were rior insula, and ventral striatum are recruited in non-musicians
associated with the presence of several instruments playing simulta- while processing pleasant music, in contrast to unpleasant highly
neously. This finding appears to be in line with those discussed by atonal counterparts (which were created electronically by pitch-
Levitin and Menon (2003) regarding the deactivations in the posterior shifting thereby causing them to sound very unnatural or far
cingulate cortex in response to taxing auditory and visual cognitive from ‘real world’ music). Based on our findings we could postulate
tasks. Activations in the right putamen and the bilateral precentral that the participants found the tonally less clear parts of the stim-
gyrus as well as the right medial frontal gyrus, which appear to be re- ulus to be more pleasant than the tonally more clear parts. This
lated to movement (Grahn and Rowe, 2009), were found to correlate finding could be explained in light of the theory advocated by
positively with Brightness and Activity, respectively. Uddin et al. Meyer (1956) according to which one important device for elicit-
(2009) showed that these same motor-control areas are functionally ing pleasurable emotions in music listening is the tension evoked
connected to the anticorrelated DMN associated with the posterior by the violation of expectations, such as when a key is unclear
cingulate cortex. Further research, however, is called for to clarify the due to the presence of complex harmonies (Sloboda, 1991). How-
link between Brightness and motor-related brain activity. ever, more studies with an ecological listening setting are needed
In addition to the DMN, negative correlations for timbral features to clarify the relation between key clarity and emotion induction relat-
of Activity and Fullness were observed in the somatosensory areas ed to violated expectation. Furthermore, the network of these brain
(BA 2, 3 and 40). Previous studies investigating the mirror neuron areas in conjunction with activations in the medial frontal brain regions
system (Keysers et al., 2010; Koelsch et al., 2006) found that the pri- (BA 9), has been previously postulated to be involved in the processing
mary somatosensory (BA 2) and secondary somatosensory areas of affect-related evaluation and classification in music (Khalfa et al.,
were recruited while hearing sounds resulting from other people's ac- 2005) and with internal self-monitoring and evaluative processing, es-
tions. As nine out of the eleven participants reported playing the pecially in an aesthetic context. For instance, activity in this region was
piano as their primary or secondary instrument, one could postulate observed when asking subjects to judge the beauty of a painting or a
that listening to the piano parts may have activated somatosensory black-and-white abstract shape or even a rhythmic drum sequence
areas that would be active during actual piano playing. (Jacobsen et al., 2006; Kornysheva et al., 2010). These results further
Interestingly, areas of the cerebellum, including lobule VI, Crus I and hint at a link between key processing and aesthetic appreciation.
II, were found to be involved in processing timbre-related acoustic com- In line with our hypothesis, for the rhythmic component, Pulse
ponents. According to a meta-analysis by Stoodley and Schmahmann Clarity, we found correlations in the auditory cortices (BA 22 and
(2009), these areas are known to be the cognitive regions of the cere- 41), the motor cortex, basal ganglia structures (putamen), and several
bellum. Additionally, the involvement of Crus I and II in conjunction regions of the limbic system (cingulate gyrus, insula) and the right
with the superior parietal lobule, lateral prefrontal cortex, and dorsal ITG (BA 37). While the activations in the areas of the auditory cortex
premotor cortex, in processing high cognitive load in an auditory task displayed positive correlations, the remaining areas correlated
with chords was found by Salmi et al. (2010). As mentioned earlier, negatively.
high levels of Fullness and Activity in the stimulus were associated The negative correlation of Pulse Clarity with the putamen indi-
with high levels of auditory-related cognitive load. Hence, our results cates that it was highly activated during segments of the music with
for the first time demonstrate the role of the cerebellum in cognitive unclear pulse, suggesting the participants were internally generating
processing while listening to a naturalistic music stimulus. the pulse during those moments (Grahn and Rowe, 2009). Important-
Previously, the scarce attempts at identifying the neural struc- ly, the activations in interior brain structures belonging to the limbic
tures responsible for tonality processing have utilized chord ca- system in the vicinity of the amygdala, and the middle cingulate gyrus
dences especially composed or simple and repetitive melody lines in the vicinity of the supplementary motor area, have not been ob-
(Janata et al., 2002a, 2002b). In the present study, this implicit com- served to be active in previous studies where pulse was artificially
plex cognitive skill has for the first time been investigated using manipulated or where participants were asked to tap and extract
continuous expressive real music. Janata et al. (2002a, 2002b) iden- the pulse within complex rhythmic patterns. As Low Pulse Clarity im-
tified the rostromedial prefrontal cortex as a possible brain substrate plies low temporal predictability, the present finding is in line with
for tonal encoding. While, contrary to our expectations, we failed to that of Engel and Keller (2011), who report similar activations
find significant correlations either in that area or the hippocampus, when comparing temporally unpredictable improvizations with
we observed that the time series for key clarity negatively corre- their more predictable imitations. Blood and Zatorre (2001) found
lated with activations in the superior frontal gyrus (BA 9), previous- that limbic and reward areas are recruited during music-induced
ly related to beauty judgments of sound patterns (Kornysheva et al., chills. In addition, they emphasize the similarity of these results to
2010), and several brain areas related to somatomotor processing, those obtained from euphoria-related and pleasant emotion-related
particularly concentrated in BA 6 and 3 (the precentral and postcen- brain imaging studies. Moreover, recent evidence suggests that
tral gyri and the supplementary motor area) (see Fig. 4). these inner brain structures are recruited especially during intense
Importantly, we further found that decreasing clarity in the key pleasurable sensations to music (Salimpoor et al., 2011). In light of
of the musical stimulus activated inner encephalic structures relat- these findings, one could postulate that lack of clarity in the perceived
ed to emotion processing such as the claustrum and anterior cin- pulse causes tension, which could be pleasurable thereby resulting in
gulate (Etkin et al., 2010). Among the auditory areas, only the the activation of reward circuits in their brains. The present study
left HG and the bilateral rolandic operculum were activated during thus demonstrates for the first time a neural link between rhythm
processing music with unclear key. The rolandic operculum of the perception and activity of brain structures associated with emotions
premotor cortex is known to play an important role in speech ar- in a realistic listening condition.
ticulation and phonological rehearsal (Brown et al., 2005). Previ- Interestingly, we found that the right ITG (BA 37) to be activated
ous studies have further revealed that the rolandic operculum in during moments of Low Pulse Clarity. While the ITG (BA 37) has
conjunction with the insular cortex is known to play an important been associated previously with visual processing, Levitin and
role in overt and covert singing (Jeffries et al., 2003). This activa- Menon (2003) found the left ITG to be more activated during scram-
tion, coupled with the ones in the precentral and postcentral bled music listening conditions in comparison to their unscrambled
gyri, support the link between spontaneous singing and key counterparts. These findings hint at the possible role of the ITG in
V. Alluri et al. / NeuroImage 59 (2012) 3677–3689 3687
range of 0–10 Hz, indicating the average frequency of these periodic- Appendix C
ities (Pampalk et al., 2002).
Fluctuation entropy: Shannon entropy of the fluctuation spec- Appendix Table C.1
trum (Pampalk et al., 2002) representing the global repartition Summary of mean inter-subject correlations within the areas correlating significantly
with each of the acoustic components. Columns two and three indicate the means and
of rhythm periodicities. Fluctuation entropy is a measure of the standard deviations of the mean inter-subject correlation values across the voxels display-
noisiness of the fluctuation spectrum. For example, a noisy fluctu- ing significant correlations with each of the acoustic components. Columns four through
ation spectrum can be indicative of several co-existing rhythms of six indicate, for the voxels correlating significantly with each acoustic component, the per-
different periodicities, thereby indicating a high level of rhythmic centage of voxels with significant mean inter-subject correlation (evaluated at p b .05,
p b .01, p b .001).
complexity.
Pulse clarity: estimate of clarity of the pulse (Lartillot et al., Inter-subject Overlap
2008). correlation
1 2 X N−j
Fullness Brightness Timbral Key Pulse
1
≈ þ ρ ðjÞρYY ðjÞ B:1 Complexity Clarity Clarity
df N N j N XX
Brightness .51⁎⁎
Timbral complexity .33⁎ −.21
Key clarity −.11 −.12 .02
where ρXX(j) is the normalized autocorrelation of the signal of N
Pulse clarity .49⁎⁎ .24 −.26 −.15
observations at lag j. The maximal lag j chosen was N/5 because Activity .92⁎⁎⁎ .64⁎⁎⁎ −.27 −.12 .50⁎⁎
it is known to yield relatively accurate results in terms of error
⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.
rates (Pyper and Peterman, 1998).
Fox, M.D., Zhang, D., Snyder, A.Z., Raichle, M.E., 2009. The Global Signal and Observed Meyer, L.B., 1956. Emotion and meaning in music. University of Chicago Press, Chicago.
Anticorrelated Resting State Brain Networks. J. Neurophysiol. 101 (6), 3270–3283. Pallesen, K.J., Brattico, E., Bailey, C.J., Korvenoja, A., Gjedde, A., 2009. Cognitive and
Fraisse, P., 1982. Rhythm and tempo. In: Deutsch, D. (Ed.), Psychol. Music. Academic emotional modulation of brain default operation. J. Cogn. Neurosci. 21 (6),
Press, New York. 1065–1080.
Friston, K.J., Josephs, O., Zarahn, E., Holmes, A.P., Rouquette, S., Poline, J.-B., 2000. To Pampalk, E., Rauber, A., Merkl, D., 2002. Content-based organization and visualization
smooth or not to smooth? Neuroimage 12, 196–208. of music archives. Proc. IEEE ACM-MM, pp. 570–579.
Gómez, E., 2006. Tonal description of music audio signal. PhD Thesis, Universitat Pom- Pantev, C., Roberts, L.E., Schulz, M., Engelien, A., Ross, B., 2001. Timbre-specific en-
peu Fabra, Barcelona. hancement of auditory cortical representations in musicians. Neuroreport 12,
Grahn, J.A., Rowe, J.B., 2009. Feeling the beat: premotor and striatal interactions in mu- 169–174.
sicians and nonmusicians during beat perception. J. Neurosci. 29, 7540–7548. Patterson, R.D., Uppenkamp, S., Johnsrude, I.S., Griffiths, T.D., 2002. The processing
Hahn, B., Ross, T.J., Stein, E.A., 2006. Neuroanatomical dissociation between bottom-up of temporal pitch and melody information in auditory cortex. Neuron 36,
and top-down processes of visuospatial selective attention. Neuroimage 32, 842–853. 767–776.
Halpern, A.R., Zatorre, R.J., Bouffard, M., Johnson, J.A., 2004. Behavioral and neural corre- Pyper, B.J., Peterman, R.M., 1998. Comparison of methods to account for autocorrela-
lates of perceived and imagined musical timbre. Neuropsychologia 42, 1281–1292. tion in correlation analyses of fish data. Can. J. Fish. Aquat. Sci. 55, 2127–2140.
Harrington, D.L., Haaland, K., Hermanowicz, N., 1998. Temporal processing in the basal Rao, S.M., Mayer, A.R., Harrington, D.L., 2001. The evolution of brain activation during
ganglia. Neuropsychology 12, 3–12. temporal processing. Nat. Neurosci. 4, 317–323.
Hasson, U., Nir, Y., Levy, I., Fuhrmann, G., Malach, R., 2004. Intersubject synchronization Saari, P., Eerola, T., Lartillot, O., 2011. Generalizability and Simplicity as Criteria in
of cortical activity during natural vision. Science 303, 1634–1640. Feature Selection: Application to Mood Classification in Music. IEEE Transactions
He, B.J., Snyder, A.Z., Zempel, J.M., Smyth, M.D., Raichle, M.E., 2008. Electrophysiological on Audio, Speech, and Language Processing 19 (6), 1802–1812. doi:10.1109/
correlates of the brain's intrinsic large-scale functional architecture. Proc. Natl. TASL.2010.2101596.
Acad. Sci. 105, 16039–16044. Salimpoor, V., Benovoy, M., Larcher, K., Dagher, A., Zatorre, R.J., 2011. Anatomically dis-
Jacobsen, T., Schubotz, R.I., Hofel, L., Cramon, D.Y., 2006. Brain correlates of aesthetic tinct dopamine release during anticipation and experience of peak emotion to
judgment of beauty. Neuroimage 29 (1), 276–285. music. Nat. Neurosci. 14 (2), 257–262.
Janata, P., 2005. Brain networks that track musical structure. Ann. N. Y. Acad. Sci. 1060, Salmi, J., Pallesen, K.J., Neuvonen, T., Brattico, E., Korvenoja, A., Salonen, O., et al., 2010.
111–124. Cognitive and motor loops of the human cerebro-cerebellar system. J. Cogn. Neu-
Janata, P., Grafton, S., 2003. Swinging in the brain: shared neural substrates for behav- rosci. 22, 2663–2676.
iors related to sequencing and music. Nat. Neurosci. 6, 682–687. Samson, F., Zeffiro, A.T., Toussaint, A., Belin, P., 2011. Stimulus complexity and categor-
Janata, P., Birk, J.L., van Horn, J.D., Leman, M., Tillmann, B., Bharucha, J.J., 2002a. The cortical ical effects in human auditory cortex: an activation likelihood estimation meta-
topography of tonal structures underlying western music. Science 298, 2167–2170. analysis. Front. Psychol. 1, 1–23.
Janata, P., Tillmann, B., Bharucha, J.J., 2002b. Listening to polyphonic music recruits Schaefer, R.S., Desain, P., Suppes, P., 2009. Structural decomposition of EEG signatures
domain-general attention and working memory circuits. Cogn. Affect. Behav. Neu- of melodic processing. Biol. Psychol. 82, 253–259.
rosci. 2, 121–140. Schubotz, R.I., von Cramon, D.Y., 2002. Predicting perceptual events activates corre-
Jeffries, K., Fritz, J., Braun, A., 2003. Words in melody: an H215O PET study of brain acti- sponding motor schemes in lateral premotor cortex: an fMRI study. Neuroimage
vation during singing and speaking. Neuroreport 14, 749–754. 15, 787–796.
Keysers, C., Kaas, J.H., Gazzola, V., 2010. Somatosensation in social perception. Nat. Rev. Schwartze, M., Keller, P.E., Patel, A.D., Kotz, S.A., 2011. The impact of basal ganglia
Neurosci. 11, 417–428. lesions on sensorimotor synchronization, spontaneous motor tempo, and the
Khalfa, S., Schon, D., Anton, J.L., Liegeois-Chauvel, C., 2005. Brain regions involved in the detection of tempo changes. Behav. Brain Res. 216 (2), 685–691.
recognition of happiness and sadness in music. Neuroreport 16, 1981–1984. Sethares, W.A., 1998. Tuning, Timbre, Spectrum, Scale. Springer-Verlag.
Koelsch, S., Fritz, T., von Cramon, D.Y., Muller, K., Friederici, A.D., 2006. Investigating Shannon, C.E., 1948. A mathematical theory of communication. Bell Syst. Tech. J. 27,
emotion with music: an fMRI study. Hum. Brain Mapp. 27, 239–250. 379–423.
Kornysheva, K., von Cramon, D.Y., Jacobsen, T., Schubotz, R.I., 2010. Tuning-in to the Sloboda, J., 1991. Music structure and emotional response: some empirical findings.
beat: aesthetic appreciation of musical rhythms correlates with a premotor activity Psychol. Music. 19, 110–120.
boost. Hum. Brain Mapp. 31 (1), 48–64. Smith, A.M., Lewis, B.K., Ruttimann, U.E., Ye, F.Q., Sinnwell, T.M., Yang, Y., Duyn, J.H.,
Krumhansl, C.L., 1990. Cognitive Foundations of Musical Pitch. Oxford University Press, Frank, J.A., 1999. Investigation of low frequency drift in fMRI signal. Neuroimage
Oxford. 9, 526–533.
Lartillot, O., Toiviainen, P., 2007. MIR in Matlab (II): a toolbox for musical feature ex- Stoodley, C.J., Schmahmann, J.D., 2009. Functional topography in the human cerebel-
traction from audio. In: Dixon, S., Bainbridge, D., Typke, Rainer (Eds.), Proc. Intl. lum: a meta-analysis of neuroimaging studies. Neuroimage 44, 489–501.
Conf. Music Inform. Retrieval, pp. 237–244. Tzanetakis, G., Cook, P., 2002. Music genre classification of audio signals. Proc. IEEE T
Lartillot, O., Eerola, T., Toiviainen, P., Fornari, J., 2008. Multi-feature modeling of pulse Acoust Speech. 10, pp. 293–302.
clarity: design, validation, and optimization. Proc. ISMIR, Philadelphia. Uddin, L.Q., Kelly, A.M.C., Biswal, B.B., Xavier Castellanos, F., Milham, M.P., 2009. Func-
Lazar, N.A., 2008. The statistical analysis of functional MRI data. Springer, New York. tional connectivity of default mode network components: correlation, anticorrela-
Ledberg, A., Akerman, S., Roland, P.E., 1998. Estimation of the probabilities of 3D clus- tion, and causality. Hum. Brain Mapp. 30, 625–637.
ters in functional brain images. Neuroimage 8, 113–128. Wong, P.C., Skoe, E., Russo, N.M., Dees, T., Kraus, N., 2007. Musical experience shapes
Levitin, D.J., Menon, V., 2003. Musical structure is processed in “language” areas of the human brainstem encoding of linguistic pitch patterns. Nat. Neurosci. 10,
brain: a possible role for Brodmann area 47 in temporal coherence. Neuroimage 420–422.
20, 2142–2152. Zanto, T.P., Snyder, J.S., Large, E.W., 2006. Neural correlates of rhythmic expectancy.
Mcavoy, M., Larson-Prior, L., Nolan, T.S., Vaishnavi, S.N., Raichle, M.E., d'Avossa, G., Adv. Cogn. Psychol. 2, 221–231.
2008. Resting states affect spontaneous BOLD oscillations in sensory and paralim- Zatorre, R.J., Belin, P., Penhune, V.B., 2002. Structure and function of auditory cortex:
bic cortex. J. Neurophysiol. 100 (2), 922–931. music and speech. Trends Cogn. Sci. 6, 37–46.